Göran Krampe wrote:
It looks a bit wordy but, what do you think?
Well, honestly I don't like XML much. :)
But either we pick ONE single format (for simplicity) and then I am leaning towards a simple chunk format very much like previous formats (.st .cs). You know - KISS.
XML is not necessarily a bad option for this goal. We have an XML parser that works very well, is relatively small and fast and allows for all but the most esoteric stuff. Chunks are okay but the disadvantage with them is that they are context-sensitive and that a parser can therefore not ignore something it doesn't understand which makes extensibility a brittle thing. A good example is that a 3.8 file contents browser will be unable to correctly parse fileOuts created from 3.9 - that's because it expects to find a class definition being made up of a particular number of tokens and that number of tokens has changed in 3.9 (which is impossible to tell without some sort of meta information). However, if that problem had been addressed by another (context-sensitive) chunk it would probably have exploded the FCB altogether ;-)
Since I can't do all work I will focus on the simple chunk-format but will take into account that the system may have multiple "parsers" to use when reading/writing Deltas. Thus we can add more formats as we go. If you are interested in helping with an XML-format, that would be nice! :)
I'll check it out. Unfortunately, I'm completely swamped with work right now.
Cheers, - Andreas